espnet2.uasr.discriminator.abs_discriminator.AbsDiscriminator

source

class espnet2.uasr.discriminator.abs_discriminator.AbsDiscriminator(*args, **kwargs)

Bases: Module, ABC

Abstract base class for implementing discriminators in the ESPnet2 framework.

This class defines the interface for all discriminator implementations. It inherits from torch.nn.Module and requires subclasses to implement the forward method. The forward method is responsible for processing input tensors and producing output tensors, typically used in adversarial training settings.

None

• Parameters:
  • xs_pad (torch.Tensor) – A padded input tensor containing the features.
  • padding_mask (torch.Tensor) – A tensor indicating the padding positions in xs_pad.
• Returns: The output tensor produced by the discriminator after : processing the input features.
• Return type: torch.Tensor
• Raises:NotImplementedError – If the forward method is not implemented in a subclass.

####### Examples

To create a custom discriminator, inherit from this class and implement the forward method as follows:

``

`

python class MyDiscriminator(AbsDiscriminator):

def forward(self, xs_pad, padding_mask): : # Custom processing logic here return output_tensor

``

`

NOTE

This class is intended to be subclassed, and should not be instantiated directly.

Initialize internal Module state, shared by both nn.Module and ScriptModule.

abstract forward(xs_pad: Tensor, padding_mask: Tensor) → Tensor

Computes the forward pass of the discriminator model.

This method takes padded input sequences and their corresponding padding masks to produce a tensor output. It must be implemented by any subclass of AbsDiscriminator.

xs_pad

Padded input sequences for the discriminator.

• Type: torch.Tensor

padding_mask

Mask indicating the positions of padding in

• Type: torch.Tensor

the input sequences.

• Parameters:
  • xs_pad (torch.Tensor) – A tensor of shape (batch_size, seq_length,
  • data. (feature_dim ) containing the padded input)
  • padding_mask (torch.Tensor) – A tensor of shape (batch_size, seq_length)
  • valid (indicating which elements of xs_pad are)
• Returns: A tensor containing the output from the discriminator, with shape (batch_size, output_dim).
• Return type: torch.Tensor
• Raises:NotImplementedError – If the method is not implemented in a subclass.

####### Examples

>>> discriminator = MyDiscriminator()  # MyDiscriminator should extend AbsDiscriminator
>>> xs_pad = torch.rand(32, 10, 64)  # Example input (32 samples, seq_len=10, features=64)
>>> padding_mask = torch.ones(32, 10)  # Example mask (no padding)
>>> output = discriminator(xs_pad, padding_mask)
>>> print(output.shape)  # Output shape will depend on the implementation

NOTE

This method must be overridden in any subclass of AbsDiscriminator.